Analysis of highway traffic control of human and autonomous driving
Traffic jam on the highway could waste people’s time, reduce fuel efficiency and increase air pollution. When people are stuck in traffic jam, it is intuitive to believe there are major collisions ahead. However, from analysis, traffic jams could happen without bottlenecks when the vehicle density exceeds some critical density. The cause of the traffic jam is due to the pattern that people drive, represented as Car-following Model. In the Car-following Model, people adjust the relative distance and speed with respect to the leading cars only. By eigenvalue and eigenvector analysis, this model is unstable in face of small perturbations of the relative space or speed between cars.
While it is difficult to change people’s driving habits, the emergence of self driving car makes it possible to design new traffic systems. In recent research, it is proposed that bilateral control model can be used to suppress instabilities in traffic flow. In this model, the state of a car is determined by the states of both leading and following cars, which can be achieved by sensor inputs, e.g. radar or lidar. Since the sensors provide more information about the surrounding and the computations could be done faster than human reactions, self driving cars present an opportunity to smooth traffic jams.